What drives people? Analyzing leisure-shopping trip decision making

(1)

What Drives People?

1

Analyzing Leisure-shopping Trip Decision Making

2 3 4 5 6 7 8 9 10 Authors: 11 12

Tim De Ceunynck, Diana Kusumastuti, Els Hannes, Davy Janssens and 13

Geert Wets (corresponding author) 14

Transportation Research Institute, Hasselt University 15

Wetenschapspark 5/6, 3590 Diepenbeek, Belgium 16

Fax: +32(0)11 26 91 99 17

Tel.: +32(0)11 26 -- -- {91 23; 91 23; 91 34; 91 28; 91 58} 18

E-mail: {tim.deceunynck; diana.kusumastuti; els.hannes; davy.janssens; 19

geert.wets}@uhasselt.be 20

21 22

First submitted: 30 July 2010

23

Revised paper submitted: 12 November 2010

24 25 26 27 28 29 30 31 32 33 34 35 36 Word counts: 37 Abstract 192 38 Body of paper 4962 39 Figures (NF*250) 1*250 250 40 Tables (NT*250) 5*250 1250 41 Photographs (NP*250) 0*250 0 42 References 735 43 Total 7389 44 45 46

(2)

ABSTRACT

1

2

Because of the strong increase in the number of leisure-shopping trips, a shift towards more 3

sustainable leisure-shopping behaviour is desirable. This can be attained by having a better 4

insight into people’s reasoning in choosing a transport mode and shopping location for this 5

type of activities. Thus, this paper highlights individuals’ leisure-trip decision-making 6

processes. The uniqueness of this study is the use of a large sample group, consisting of 221 7

respondents. A Computer-Based Causal Network Elicitation Technique is developed for this 8

purpose, and participants’ responses are analyzed by means of Frequent Itemset Analysis. It 9

appears that the complexity of the mental representation of the decision problem is very 10

stable over different socio-demographic groups. However, clear differences appear between 11

these groups concerning the content of the mental representation. The most remarkable 12

findings are the limited importance of cost and environmental aspects in the transport mode 13

choice. This has important implications for policy and marketing efforts to encourage 14

sustainable transport modes for leisure-shopping. It is recommended to focus advertising 15

campaigns and policy measures on aspects that are most important in people’s decision-16

making process: flexibility, travel time, accessibility, easiness for parking and some practical 17

concerns. 18

(3)

INTRODUCTION

1 2

In the last decades there has been a prominent growth of the yearly number of person car 3

kilometres in Belgium (1), the country where this research is set, resulting in the upsurge of 4

its negative externalities such as car crashes, emissions, congestion, etc., as can be witnessed 5

worldwide. The increasing importance of leisure activities, including leisure-shopping, 6

contributes to this rise (2). Most leisure-shopping trips in Belgium are oriented towards city 7

centres where the majority of stores are located, creating abundant traffic there and exposing 8

many people to its negative influences (3). Furthermore, both driving and parked cars take up 9

a lot of valuable space in this area that can be put to better use. Considering the importance of 10

leisure activities, especially in developed countries, and their influences on the increased use 11

of cars, this paper focuses on examining people’s leisure-shopping travel behaviour in the 12

city centre. 13

14

Therefore, a shift towards more sustainable leisure-shopping behaviour is required to obtain a 15

more sustainable transportation system. Travel Demand Management (TDM) measures are 16

commonly used for this purpose. These measures reduce travel demand, or redistribute this 17

demand in space, in time or by transport mode to reduce its negative impacts (4). 18

Accordingly, different transport mode characteristics are influenced to boost the competitive 19

position of sustainable transport modalities compared to cars. TDM measures can be 20

implemented more effectively if they are aimed at aspects that have a strong influence in 21

people’s decision-making. This means that travel choices should be studied at a disaggregate 22

level, as the outcome of each individual’s decision process (5). Results of such studies may 23

give better behavioural insights that can be used as feedback for policy makers to develop 24

high impact policy measures. 25

26

Thus, this study investigates people’s decision making processes when carrying out leisure-27

shopping activities using the Causal Network Elicitation Technique (CNET). CNET is a 28

qualitative research method to elicit individuals’ constructs and beliefs when making travel 29

decisions, and the links between them in a structured mental representation (MR) of the 30

decision problem (5). Originally, CNET is developed as a semi-structured interview 31

technique. Participants are asked to think aloud when resolving certain problem tasks, and 32

listing all considerations that come to their mind. Based on the elicited variables, interviewers 33

have to ask further questions until the underlying benefits that individuals want to gain are 34

revealed. 35

36

In this study, the technique is transferred to a fully automated computer-based (CB) survey 37

named CB-CNET. It works by showing participants some predefined variables and asking 38

them to sort out all relevant aspects that appear in their thought processes. A CB survey 39

offers multiple cost and methodological advantages over the original face-to-face technique. 40

These will be discussed into more detail later in this paper. 41

42

All respondents live in the vicinity of the city centre of Hasselt, a city in Belgium with a 43

typical European historical city centre. Participants’ transport mode and shopping location 44

decision for leisure-shopping activities are investigated into detail. Frequent itemset analysis 45

(4)

is applied to reveal frequently elicited variables in respondents’ MR, and important 1

associations between the different variables. 2

3

This paper is structured as follows. First, a general explanation about different types of 4

decision-making is presented. Next, the concept of MR is elucidated. Then, the structure of 5

the CB-CNET survey is briefly explained. Subsequently, frequent itemset analysis is 6

clarified. Next, the results of the analyses are presented. Finally, the paper ends with a 7

discussion and policy implications section. 8

9

DECISION-MAKING TYPES

10 11

In this section, a brief introduction about human decision-making is presented. This allows 12

the reader to understand the different structures that are present in people’s MR, and that are 13

elicited by the CB-CNET. 14

15

One of the most influential theories that tries to explain people’s decision-making process is 16

rational choice theory. It assumes that people calculate the likely costs and benefits of any 17

action before deciding what to do (6). The decision maker is faced with a set of alternative 18

choice options, of which he is assumed to choose the alternative that yields the highest 19

expected utility (7). Although the theory offers valuable insights, it is also often criticized 20

because of its unrealistic assumptions (e.g. 10). People are often not fully informed; they are 21

not always completely self-interested (9; 10); and some decisions are not only based on 22

reason but also on emotions or intuition (11; 12). Therefore, people’s decision-making 23

process can be seen as a process relying on a number of simplifying heuristics, rather than 24

extensive algorithmic processing (13). Heuristics are efficient rules of thumb of the type if-25

then(-else) to get to a decision relatively easily. They are based on experiences and 26

knowledge (8). Especially when decisions are repeated many times, like in daily travel, 27

people do not go through the whole complex decision procedure each time. In this case, 28

heuristic decisions are commonly used to reduce the mental effort that is required for 29

weighing and judging the possible decision alternatives (14). 30

31

However, for new or occasional decisions, it is possible that people do not have ready-made 32

solutions for all possible occurring contexts in the decision environment. In this case, a 33

rational decision-making process can be activated. Research shows that, in case of a new or 34

infrequently occurring decision problem, the decision maker activates a complex and 35

deliberative cognitive process to make the best decision (15). In this decision process, 36

different considerations are linked by means of causal relations (14). This way, a temporary 37

MR of the decision problem is created in which the decision maker details relevant attributes 38

of the different alternatives and judges their subjective values, attractiveness or suitability 39

(16). 40

41

Besides rational and heuristic decision-making, there is habitual (or automated) behaviour, 42

that develops as people repeat actions in stable circumstances. When initially performing an 43

action, people consciously decide what to do and how to do it in order to achieve certain 44

outcomes. As people repeat actions, the conscious decision-making process recedes, and the 45

actions come to be triggered by the environment (17). 46

(5)

1

Of course, there is a large twilight zone between these extremes of fully conscious and 2

deliberate decisions, and fully automated habitual responses. Leisure-shopping belongs to this 3

twilight zone. For most people, leisure-shopping is not an activity that is executed daily, or 4

even weekly. Furthermore, it is subject to an ever-changing context: an interest in a specific 5

product, the weather, etc. Therefore, leisure-shopping related decisions will not be fully 6

automated for most people. On the other hand, if leisure-shopping is executed occasionally, 7

some aspects of leisure-shopping decision-making can become automated to some extent. 8

That is why people’s MR about leisure-shopping decisions may contain both elements of 9

conscious deliberation and heuristics. 10

11

MENTAL REPRESENTATIONS

12 13

It is argued in literature (5; 16) that people activate a MR when solving certain problem tasks, 14

especially when facing complex decision problems and when a rational decision-making style 15

is activated. A MR consists of various arisen contexts in the decision environment, the benefit 16

requirements, instruments of the shopping trip’s decision alternatives, and the causal 17

relationships between these variables (16). Thus, four types of variables can be distinguished 18

in the MR: decision, contextual, instrumental and benefit variables (18), and they are 19

discussed respectively in the following paragraphs. Eliciting individuals’ MRs can also help 20

to understand the underlying reasoning behind decision choices using heuristic decision-21

making and habitual styles. By knowing the underlying benefits, contexts and instruments 22

that people want to gain, high impact TDMs can be selected more effectively. 23

24

Decision variables represent the decision alternatives available to the decision maker. For

25

each decision variable, there is a set of pre-defined choice alternatives (19), such as using car, 26

bus or bicycle to go to the city centre. 27

28

The different choice alternatives have various characteristics, leading to different 29

consequences. The particular characteristics of choice alternatives can be considerations in 30

the decisions. Such considerations are called instrumental variables (20). Instrumental 31

aspects can be observed and operated by the decision maker (5). Examples of instruments of 32

different transport mode options are speed, travel time, etc. 33

(6)

1

FIGURE 1 An example of the different variable types and their links in a MR (15). 2

3

Contextual variables refer to given circumstances, situations and constraints in the decision

4

environment that influence the outcome of a decision, but that cannot be controlled by the 5

decision maker (15; 19). These can be natural forces (e.g. weather conditions), and different 6

constraints, i.e. capability, coupling and authority constraints (21). 7

8

Benefit variables are directly related to utilities (20). They describe the impact of the state of

9

the contextual and instrumental variables on the fundamental needs of the decision maker and 10

his well-being (16). An example of a benefit variable is the desire to gain efficiency, physical 11

comfort, etc. 12

13

The last elements in the MR are the causal links between contextual, instrumental and 14

benefit variables to obtain a network representation of the decision problem. Hence, this 15

(7)

causal network represents the individual’s beliefs about how different variables activate the 1

consideration of other decision-related variables (18). 2

3

The smallest building block of a MR of a decision problem is called a “cognitive subset”. A 4

cognitive subset is a connected set of considerations, consisting of a context, a benefit and an 5

instrument (5). However, it is also possible that a combination of a benefit and an instrument 6

is always considered, irrespective of the circumstances, for instance in the habitual decision 7

making style. Therefore, the second type of cognitive subset applicable in all circumstances is 8

registered; i.e. (normally)-benefit-instrument (15). 9

10

An example of the different elements in a person’s MR related to leisure-shopping trip 11

decision-making is shown in the Figure 1. 12

13

METHODOLOGY: COMPUTER-BASED CNET SURVEY

14 15

Initially, CNET is developed as a semi-structured face-to-face interview technique. In the 16

CNET interview protocol, the interviewer asks a series of questions, and checks the 17

respondent’s answers with a pre-defined list of variables. Later, the CNET card game method 18

has been developed, which makes use of written cues instead of spontaneous recalling. 19

Written cues give better results than spontaneously recalling because they reduce the chance 20

that respondents forget to mention relevant aspects (22; 23). Interested readers are referred to 21

(5; 18) for research using these techniques. For this study, it is decided to transfer the CNET 22

card game technique to a computer-based survey. Compared to both other techniques, a 23

computer-based survey offers important benefits; i.e. cutting the marginal costs and 24

eliminating interviewer’s biases (24). Moreover, the CB-CNET enables gathering the leisure-25

shopping MR data from a large sample group, which is practically infeasible using both other 26

techniques. 27

28

The CB-CNET is divided into several parts. In the first part, respondents are asked for their 29

personal information. Socio-demographic factors like age, gender, income, education, etc. are 30

questioned. This information is used to investigate differences between different subgroups of 31

the sample, and to check the representativeness of the sample. 32

33

In the second part, the leisure-shopping setting is explained. Respondents are asked to 34

imagine that they want to buy something (a gift and/or something to wear) for an upcoming 35

party of a friend. Next, they are asked to give their considerations related to their transport 36

mode (TM) choice and their shopping location (SL) choice, i.e.: the shopping zone in Hasselt 37

they will go to first. Concerning the SL choice, the city centre is subdivided into three zones: 38

the main shopping street, the boutique area, and the gallery area. The TM decision assumes 39

that all respondents have a uniform choice set consisting of bus, bicycle and car. This 40

assumption is valid since an analysis of the data has shown that all-but-one respondents have 41

a car in their household, more than 90% own a bicycle, and the Decree on Basic Mobility 42

guarantees a bus stop within 750m to all inhabitants of residential areas (25). 43

44

In the third step, respondents are asked which of both decisions they will take first and which 45

one last. Then, they are asked to indicate whether their decision-making for this decision 46

(8)

depends on certain circumstances, indicating heuristic or rational decision-making, or 1

whether this choice can be made spontaneously, which indicates habitual decision-making. 2

This is called the split-elicitation procedure: based on the respondent’s answer, different 3

elicitation paths are followed (15). Both paths are analogous for both decisions. 4

5

In case the respondent indicates that his decision depends on certain circumstances, the 6

elicitation starts with the contextual variables. For each contextual variable, one or more 7

benefit variables can be indicated, and for each context-benefit combination, one or more 8

instrumental variables can be chosen. Respondents are asked to limit their responses to the 9

most important contexts and benefits to keep the burden of the survey to an acceptable level. 10

A shortened list is shown for the elicitation of the instrumental aspects, containing only 11

relevant instruments that have been identified from literature, pilot tests (5; 18) and common 12

sense. To ensure that all respondents have the same interpretation of the variables, the 13

definition of a variable is shown in the interface. As soon as the elicitation of these context-14

specific cognitive subsets is finished, the respondent has the opportunity to indicate not 15

context-specific cognitive subsets. However, this is not mandatory. 16

17

On the other hand, if the respondent indicates that he can make the decision straightaway, the 18

elicitation process reverses. Respondents are initially asked to elicit the (normally)-benefit-19

instrument subsets, and afterwards, a chance is given to indicate subsets of the type context-20

benefit-instrument. The elicitation procedure ends with an open-ended question to reveal 21

additional and important considerations that are not presented in the lists. The results of this 22

question will be used as input to improve the list of variables in future research. 23

24

Respondent requirements, i.e. living on the outskirts of Hasselt, possessing a driver’s license, 25

and a high willingness to cooperate in the research, limit the sample in this research. 26

Therefore, the sample is gathered by means of the snowball sampling method. This method is 27

mostly used in case of a rare sample characteristic (26). Because of the length of the survey 28

(approximately two hours), the 221 respondents fill out the survey in small guided group 29

sessions of up to 16 persons per session, reducing the risk of unreliable data. 30

31

The lists of variables are formulated based on an extensive literature review and pilot studies 32

(5; 18). The list of variables for the SL decision consists of 15 contextual, 15 benefit and 22 33

instrumental variables. The list of variables for the TM decision consists of 27 contextual, 15 34

benefit and 25 instrumental variables. 35

36

ANALYSIS

37 38

Frequent Itemset Analysis

39 40

The data gathering procedure results in an extensive dataset of 221 respondents, who have 41

elicited 12701 cognitive subsets. This should be analyzed in an efficient way. Therefore, a 42

suitable data mining technique, i.e. frequent itemset analysis, is required to identify not only 43

the frequent aspects, but also the associations between them. Thus, this technique is suited to 44

capture important cognitive subsets in participants’ MRs. 45

(9)

A frequent itemset is a combination of items that frequently appears in the dataset (5). What 1

itemsets are exactly considered to be “frequent” is usually determined by a threshold value, 2

which is called the minimal support value (minsupp). If the support value of a certain 3

combination of items is equal to or higher than the minsupp value, the itemset is considered 4

as a frequent itemset. The support value is usually expressed as a percentage of the total 5

number of transactions in the database that contains the combination of items (27). In this 6

study, frequent itemset analysis is used as a tool to describe frequent patterns of cognitive 7

subsets generated from the elicited data. For a more detailed explanation about frequent 8

itemset analysis, the reader is referred to (27). 9

10

An adjusted form of the support value is used in this study because each cognitive subset is 11

coded as one record in the database, and each respondent can only elicit each cognitive subset 12

once. Hence, the database consists of multiple records for each respondent. This results in a 13

very low expected support value for each cognitive subset, and the value itself becomes 14

difficult to interpret. Therefore, the support value is adjusted so that it indicates the 15

percentage of respondents who have indicated the particular subset. This value is called the 16

compensated support value (csupp). An example will clarify this problem. Suppose the 17

following database, consisting of two respondents each eliciting 5 cognitive subsets: 18 19 Respondent 1: 20 (normally)-efficiency-flexibility 21 (normally)-freedom-travel time 22 (normally)-freedom-accessibility 23 (normally)-convenience-flexibility 24 precipitation-physical comfort-shelter 25 Respondent 2: 26 (normally)-efficiency-flexibility 27 (normally)-freedom-flexibility 28 (normally)-cost-saving money 29 precipitation-physical comfort-shelter 30

time available-efficiency-travel time 31

32

The support value for the cognitive subset “(normally)-efficiency-flexibility” is 0,20 (i.e., 33

20% of the database contains this combination of items). However, since each respondent 34

delivers multiple data entries, the value is difficult to interpret. The csupp value compensates 35

for the number of data entries per person. Therefore, the csupp value for this cognitive subset 36

is 1,00 (i.e., 100% of respondents in the database has elicited this cognitive subset). 37

38

One fixed minsupp value cannot be used for all analyses because the dataset has to be split 39

into several sets based on socio-demographic characteristics and other measures, resulting in 40

new datasets that may contain only a small number of records. Thus, one fixed minsupp value 41

can be too low for some analyses or too high for others. Therefore the five most frequent 42

itemsets in each analysis are used to discuss the results. Furthermore, the context-specific and 43

the not context-specific frequent itemsets are analyzed separately. This is done because they 44

are difficult to compare, since, purely from a probabilistic point of view, the chance that one 45

particular set consisting of three elements is selected is smaller than the chance that one 46

(10)

particular set of two elements is selected. Preliminary analyses have shown that analyzing 1

them together results in a list of frequent itemsets mainly consisting of the subsets of the 2

“normally-value-instrument” type. So, if this split is not made, it is difficult to draw 3

conclusions about the influence of contextual variables. 4 5 RESULTS 6 7 Complexity of Respondents’ MR 8 9

The number of elicited variables in each participant’s MR is calculated as a complexity 10

measure. First, the MR complexity of different socio-demographic groups is investigated. For 11

socio-demographic groups who elicit a more complex MR, each element or association has a 12

relatively lower importance on average, because of the large number of aspects taken into 13

account. Furthermore, the chance that each element or association present in the MR is higher 14

for groups who elicit a more complex network. In case the analysis shows that certain socio-15

demographic groups have a more complex MR than others, it is checked if these groups are 16

underrepresented or overrepresented in the sample. The results are presented in Table 1. 17

18

It can be concluded that the network complexity is quite constant among different socio-19

demographic groups. Respondents indicate on average about 44 different variables. The TM 20

decision (27 variables) is significantly more complex than the SL decision (17 variables). 21

There are large individual differences in the complexity of the MR; the simplest MR contains 22

only 10 different variables, while the most complex one consists of 104 variables. However, 23

despite these large individual differences, averages are very stable across the different socio-24

demographic groups. Only education level seems to have an influence on the MR complexity. 25

Higher educated respondents indicate significantly less variables than lower educated ones 26

for both decisions. This may happen because participants are explicitly asked to limit the 27

elicitation to the most important variables for their decision. Higher educated respondents are 28

probably more able to distinguish the importance of the variables, resulting in a lower 29

number of elicited variables. Further research is needed to elucidate this issue. 30

(11)

TABLE 1 Complexity of Respondents’ MR.

1

Subgroup of sample No. of respondents Network complexity:

Best estimate [95% confidence interval] Overall 221 44,199 [41,961; 46,437] SL decision 221 16,742 [15,688; 17,796] * TM decision 221 27,457 [25,970; 28,944] * Gender Men 95 44,516 [41,087; 47,945] Women 126 43,960 [40,994; 46,926] Age 19-29 54 43,426 [39,644; 47,208] 30-39 25 42,160 (35,630; 48,690)** 40-49 48 43,750 [38,361; 49,139] 50-59 56 44,589 [40,009; 49,169] 60+ 38 46,632 [40,942; 52,322] Education level

Lower educated (secondary degree or lower) 83 49,458 [45,342; 53,574] *

Higher educated 138 41,036 [38,579; 43,493] *

Net household income category

Low (max. €2000 / month) 65 48,077 [43,412; 52,742]

Medium (€2001 - €4000 / month) 93 42,000 [38,926; 45,074]

High (+ €4000 /month) 39 44,333 [37,294; 47,372]

I would rather not tell 24 42,000 (34,796; 49,204)**

Distance of residence to city centre (TM only)

Close (< 4 km) 77 26,455 [24,168; 28,742]

Medium (5 - 7 km) 88 27,875 [25,246; 30,504]

Far (8 - 10 km) 56 28,179 [25,420; 30,938]

Annual mileage as car driver (TM only)

Max. 5000 km/year 57 26,193 [23,721; 28,665]

5001 - 15000 km/year 108 28,944 [26,742; 31,146]

> 15000 km/year 54 25,056 [22,561; 27,551]

Leisure-shopping frequency

Rarely if ever 14 41,214 (32,171; 50,257)**

A few times a year 84 42,048 [38,282; 45,814]

(nearly) monthly 56 43,375 [38,941; 47,809]

A few times a month 67 48,209 [44,434; 51,984]

Note:

* There are significant differences between the different subgroups

** The confidence interval for a group with a small number of respondents (less than 30 respondents), making it difficult to draw statistically valid conclusions about the result.

2 3

(12)

Frequent Itemsets Analysis

1 2

In this section, the general findings of the frequent itemset analysis are presented and 3

analyzed for different groups; e.g. based on the split-elicitation procedure, gender, age 4

groups, education levels, income levels and distance classes from the respondent’s residence 5

to the city centre. Unfortunately, due to space limitation of the paper, only the most important 6

findings can be presented. The results of the other analyses and the definitions of the 7

variables can be obtained from the authors on request. 8

9

SL decision

10 11

Table 2 shows the results of the frequent itemset analysis for the SL choice on the full 12

database. The table shows that efficiency (saving time and effort) is a crucial aspect in the SL 13

decision, both for the context-specific frequent itemsets and the not context-specific frequent 14

itemsets. The most important instruments to gain efficiency are the presence of the 15

respondent’s favourite shop in the zone, the familiarity with the zone and the type of stores in 16

the zone. Contextual aspects that have an influence on efficiency are having an interest in a 17

specific product, and the available time to execute the leisure shopping activity. The presence 18

of the person’s favourite shop also strongly contributes to gain the benefit of fun. 19

Furthermore, the product prices in the zone have a high importance in the SL choice. 20

21

TABLE 2 Frequent Itemsets SL Decision.

22

Context Benefit Instrument Csupp*

(normally) efficiency favourite shop 28,1%

(normally) efficiency familiarity 25,8%

(normally) saving money product price 19,5%

(normally) efficiency type of stores 19,5%

(normally) fun favourite shop 18,1%

specific product efficiency favourite shop 20,8%

specific product efficiency type of stores 18,1%

time available efficiency favourite shop 17,2%

specific product efficiency familiarity 16,7%

time available efficiency familiarity 16,3%

* csupp indicates the percentage of respondents who indicated the particular subset. 23

24

TM decision

25 26

In Table 3, the results of the frequent itemset analysis for the TM choice on the full database 27

are shown. Efficiency is of crucial importance in the TM decision. The benefit “freedom” is 28

also very important in the not context-specific frequent itemsets. Both variables are often 29

related to the instruments of flexibility and travel time, and easiness for parking also 30

influences the benefit of efficiency. “Precipitation” appears to be an important contextual 31

(13)

aspect in the TM choice. This finding is in line with international literature (28). Furthermore, 1

the contextual aspects time availability, baggage and parking space availability are very 2

important in the TM decision. 3

4

A remarkable finding is that, even though financial criteria are very important in the SL 5

choice, they are considered by few people in the TM decision. Another critical finding is that 6

sustainability and the environment-friendliness of a TM are rarely considered by respondents. 7

8

TABLE 3 Frequent Itemsets TM Decision.

9

Context Benefit Instrument Csupp

(normally) freedom flexibility 39,8%

(normally) efficiency flexibility 33,9%

(normally) efficiency travel time 33,9%

(normally) efficiency easiness parking 32,1%

(normally) freedom travel time 26,7%

precipitation physical comfort shelter 21,7%

time available efficiency travel time 16,3%

baggage physical comfort treatment bags 14,9%

parking availability efficiency easiness parking 12,7%

time available efficiency flexibility 12,2%

10

In Table 4, the frequent itemsets of the TM decision are shown for people with a TM habit 11

and people without a TM habit. Subsets that are present in all categories are highlighted in 12

the same greyscale. The most apparent difference between both groups is the finding that 13

parking related aspects have a much higher importance to people without a TM habit. People 14

with a TM habit attach a relatively higher importance to associations containing the 15

contextual aspect “time availability”. Furthermore, they attach a stronger importance to the 16

instrumental aspects of flexibility and travel time. 17

18 19

(14)

TABLE 4 Frequent Itemsets TM Decision for TM habit and no TM habit.

1

TM habit

(normally) freedom travel time 39,1%

(normally) freedom accessibility 33,9%

time available efficiency accessibility 7,8%

No TM habit

(normally) efficiency easiness parking 32,1%

(normally) freedom easiness parking 24,5%

(normally) efficiency accessibility 20,8%

parking availability efficiency easiness for parking 17,9%

parking availability efficiency accessibility 14,2%

2

In Table 5, the frequent itemsets of the TM decision are shown for people from different age 3

categories. It appears that associations related to efficiency are most important to young 4

people. These are somewhat less important to elderly. Elderly attach a high importance to 5

associations with the contextual aspect “baggage” and the benefit of physical comfort. 6

Furthermore, elderly are the only age group who strongly consider financial aspects of the 7

TM choice. One final peculiar finding is that elderly strongly consider the environment-8

friendliness of the TM. 9

(15)

TABLE 5 Frequent Itemsets TM Decision for Different Age Groups.

1

Young (19-39)

(normally) efficiency easiness for parking 35,4%

(normally) efficiency accessibility 34,2%

parking availability efficiency travel time 16,5%

Middle-aged (40-59)

Elderly (60 and older)

(normally) durability environment-friendliness 26,3%

(normally) saving money cost 23,7%

time available efficiency TM preference 15,8%

baggage physical comfort accessibility 10,5%

baggage physical comfort physical effort 10,5%

2

DISCUSSION AND POLICY IMPLICATIONS

3 4

The most important policy implication of the results of the SL decision analysis is that cities 5

should make sure that their city centre is organized in an efficient way in order to make it 6

attractive for leisure-shopping. In general, efficiency can be obtained by creating walking 7

lines without detours, high density development and by grouping comparable stores together. 8

To quicken the process of getting familiar with an area, policy makers should make sure to 9

create a good legibility of their city centre (29). Since respondents attach great importance to 10

the presence of their favorite shop, cities should try to attract some popular stores to anchor 11

the MR. For instance, data shows that store chains are popular with many shoppers. Hence, 12

(16)

the presence of these stores is important to attract large shares of shoppers. An important 1

secondary goal should be to create a sufficient mix of functions and a pleasant ambience. 2

3

For policies that try to influence the TM choice, the impact of the findings in this paper are 4

quite far-reaching. These findings are clarified in the form of 5 statements. 5

6

Global warming is not so “hot” to average Joe. It appears that only few people consider the 7

environment-friendliness of the TM when making a leisure-shopping TM decision. This is 8

regrettable, since environmental concern is currently one of the “hot topics” for policy 9

makers. However, from this study it seems that the man in the street does not really share this 10

concern yet. 11

12

The ones caring most about the environment, are the ones who will not live to fully benefit 13

from the results. Remarkably enough, the only socio-demographic group that strongly 14

considers the environment-friendliness of the TM are the elderly. This is surprising, since the 15

real benefits from a sustainable TM choice can only be experienced in the long term. 16

17

Cost measures might not be worth the money. The results indicate that cost aspects are 18

considered by few respondents in their leisure-shopping TM decision. Therefore, it is 19

possible that cost measures can only accomplish limited results. Furthermore, the socio-20

demographic groups that take it most into account, are elderly and people from low income 21

households. Traditionally, these are not groups on which policy makers would want to focus 22

cost measures, because they are already weaker and more vulnerable societal groups. 23

However, it is also possible that cost considerations are not present in people’s MR because 24

transport costs are currently low. It is possible that sufficiently strong cost measures will have 25

an impact, even though financial aspects are currently not a crucial consideration in people’s 26

MR. 27

28

Policy makers should try to improve the competitive position of sustainable transport modes 29

by focusing on considerations that are really important in people’s decision-making process, 30

like flexibility, travel time, easiness for parking, accessibility, shelter provision and treatment 31

of baggage. Improving the competitive position of sustainable modes can be achieved by 32

means of pull and push measures. Pull measures are measures that stimulate the use of 33

alternative transport modes by making them more attractive. So in other words, they are soft 34

measures that try to encourage people to deliberately choose a more sustainable transport 35

mode. Push measures on the other hand are measures that aim at discouraging car use by 36

reducing its attractiveness. Hence, these are hard measures that try to force people to use a 37

more sustainable transport mode (30). 38

39

Current sustainable transport campaigns are descriptive, rather than persuasive. A 40

consequence of the finding that cost and environmental aspects are not strongly considered, is 41

that advertisement campaigns to promote sustainable TMs should be reconsidered. At this 42

moment, campaigns often aim at the fact that public transport and bicycle are cheap and 43

environment-friendly. These are in fact two of the most important advantages of these 44

sustainable TMs, so they are a good description of the TMs. But since it appears that 45

respondents do not really consider them in their TM choice, they may not be able to obtain 46

(17)

significant results. Instead, advertising campaigns to alter leisure-shopping behaviour should 1

focus more on the previously-mentioned considerations that are most important in people’s 2

decision-making process. 3

4

Parking a car in a city centre should be difficult to anyone, not only to poor drivers. One 5

particularly useful category of TDM measures to encourage sustainable TMs are parking 6

restricting measures. There are two main arguments for this. The first argument is that these 7

measures simultaneously influence several of the most important mentioned characteristics. 8

They reduce the car’s flexibility, accessibility and easiness for parking, and they can lead to 9

an increase in travel time. And second, it appears that variables related to parking are 10

particularly important to respondents without a TM habit. Nearly 50% of respondents state 11

that they do not have a TM habit for fun shopping trips, but that their choice depends on 12

circumstances. Since it is much more difficult to influence the travel behaviour of someone 13

with a TM habit than someone without a habit (17; 31), this implies that parking restricting 14

measures can have a significant impact on the modal shift of fun shopping trips. 15

16

A major downside of parking restricting measures is that the support for these measures is 17

generally low, because entrepreneurs are convinced that parking restricting measures will 18

have a negative influence on the attractiveness of the city centre. This way, they fear that they 19

will loose customers to stores and shopping centres that are located at suburban areas that are 20

easily accessible by car. However, this fear is often unjustified. Parking restricting measures 21

can increase parking comfort, make the city centre more attractive and liveable and save a lot 22

of valuable public space. This means that parking restricting measures, when they are 23

implemented well-considered and enforced strictly, can increase the attractiveness of the city 24

centre, instead of decreasing it (32). This is especially the case when the city centres offer a 25

good mix of functions, because most suburban shopping locations are mono-functional. This 26

stresses the importance of profiling the identity of the city centre. Research shows that 27

parking restricting measures lead some people to change their TM and the time of their trip, 28

but only very few will change their destination or cancel the trip (33). It is important to keep 29

in mind that this is only the case for city centres that have sufficient regional attraction. 30

31

An interesting topic for future research is to repeat the CB-CNET in other cities. This will 32

allow drawing conclusions about the generalizability of the results. It will be interesting to 33

see whether respondents from other cities have similar considerations. It will also be 34

interesting to see whether the average number of variables in the mental representation is 35

stable over different socio-demographic groups in other cities as well. Another interesting 36

topic for further research is the transferability of the method. It will be interesting to see 37

whether the CNET protocol can also be used to investigate the decision making process of 38

other decision problems. So far, it has only been used to investigate leisure shopping 39

behaviour. 40

(18)

REFERENCES

1 2

1. Belgian Federal Government. Afgelegde afstanden in het verkeer [Internet]. 2009;[cited 3

2010 Apr 2] Available from: 4

http://statbel.fgov.be/nl/statistieken/cijfers/verkeer_vervoer/verkeer/afstand/index.jsp 5

6

2. Kiiskilä M, Kalenoja H. Leisure time mobility - Future trends and sustainable 7

development. Kouvola, Finland: 2001 8

9

3. Beckx C. An activity-based modelling framework for air pollution exposure assessment. 10

2009; 11

12

4. Chen K, Miles JC. Its Handbook 2000: Recommendations from the World Road 13

Association (Piarc). 2000th ed. Artech House Publishers; 1999. 14

15

5. Kusumastuti D, Hannes E, Janssens D, Wets G, Dellaert BGC, Arentze TA. Qualitative 16

and quantitative comparisons of the CNET interview and the CNET card game to explore 17

contextual, instrumental and evaluative aspects in individuals’ fun shopping travel 18

decisions. 2009; 19

20

6. Scott J. Rational choice theory. In: Understanding contemporary society. SAGE; 2000 21

22

7. Hannes E, Janssens D, Wets G. Destination Choice in Daily Activity Travel: Mental 23

Map's Repertoire. Transportation Research Record: Journal of the Transportation 24

Research Board 2008 Dec;2054(-1):20-27. 25

26

8. Everson B, Hammer E. Thinker: Problems & Decisions: Decision Making & Heuristics 27

[Internet]. Center for the Advancement of Teaching 2005;[cited 2010 May 3] Available 28

from: http://cat.xula.edu/thinker/decisions/heuristics/ 29

30

9. Henrich J, Boyd R, Bowles S, Camerer C, Fehr E, Gintis H, McElreath R. In Search of 31

Homo Economicus: Behavioral Experiments in 15 Small-Scale Societies. The American 32

Economic Review 2001 May;91(2):73-78. 33

34

10. Kroneberg C. The Definition of the Situation and Variable Rationality: The Model of 35

Frame Selection as a General Theory of Action. Mannheim: Universität Mannheim; 36

2006. 37

38

11. Minsky M. The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and 39

the Future of the Human Mind. New York: Simon & Schuster; 2007. 40

41

12. Plessner H, Betsch C, Betsch T. Intuition in Judgment and Decision Making. 1st ed. 42

New York: Lawrence Erlbaum Associates Publishers; 2007. 43

44

13. Gilovich T, Griffin DW, Kahneman D. Heuristics and biases. Cambridge University 45

Press; 2002. 46

(19)

1

14. Kusumastuti D, Hannes E, Janssens D, Wets G, Dellaert BG. Scrutinizing Individuals’ 2

Leisure-Shopping Travel Decisions to Apprise Activity-Based Models of Travel Demand 3

(Forthcoming). Transportation, (4) 2010; 4

5

15. Kusumastuti D, Hannes E, Depaire B, Vanhoof K, Janssens D, Wets G, Dellaert BG. An 6

Interactive Computer-Based Interface to Support the Discovery of Individuals’ Mental 7

Representations and Preferences in Decisions Problems: An Application to Travel 8

Behavior. Paper is prepared for Computers in Human Behavior [date unknown]; 9

10

16. Dellaert BG, Arentze TA, Timmermans HJ. Shopping context and consumers’ mental 11

representation of complex shopping trip decision problems. Journal of Retailing 2008 12

Jun;84(2):219-232. 13

14

17. Verplanken B, Wood W. Interventions to Break and Create Consumer Habits. Journal of 15

Public Policy & Marketing 2006 May;25(1):90-103. 16

17

18. Kusumastuti D, Hannes E, Janssens D, Wets G, Dellaert BG. Qualitative exploration of 18

contextual, instrumental and evaluative aspects in individuals' fun shopping travel 19

decisions. In: Proceedings of the 12th International Conference on Travel Behavior 20

Research. Jaipur, India: 2009 21

22

19. Arentze TA, Dellaert BGC, Timmermans HJP. Modeling and Measuring Individuals' 23

Mental Representations of Complex Spatio-Temporal Decision Problems. Environment 24

and Behavior 2008 Nov;40(6):843-869. 25

26

20. den Hartog A. Activiteit en mobiliteit gevangen in een net. 2004; 27

28

21. Hägerstrand T. What about people in regional science? Papers in Regional Science 29

1970;24(1):7-21. 30

31

22. Cannell C, Oksenberg L, Converse J, editors. Experiments in interviewing techniques: 32

Field experiments in health reporting, 1971-1977. University of Michigan: Ann Arbor: 33

Survey Research Center, Institute for Social Research, University of Michigan; 1979. 34

35

23. Palmeri TJ. Automaticity. Nashville, Tennessee: 2001 36

37

24. Boardman A, Greenberg D, Vining A, Weimer D. Cost Benefit Analysis: Concepts and 38

Practice. 3rd ed. New Jersey: Pearson Prentice Hall; 2005. 39

40

25. Flemish Government. Basismobiliteit: De Basis voor een adequaat openbaar vervoernet 41

in Vlaanderen. Flemish Government, Department of Mobility and Public Works; 2008. 42

43

26. Wright R, Stein M. Snowball Sampling [Internet]. In: Encyclopedia of Social 44

Measurement. New York: Elsevier; 2005 p. 495-500.[cited 2009 Oct 20] Available from: 45

http://www.sciencedirect.com/science/article/B7GG1-4FV52TR-46

(20)

3F/2/3ffd905f68f700e73569a14808e132a4 1

2

27. Zhang C, Zhang S. Association Rule Mining: Models and Algorithms. 1st ed. Berlin, 3

Germany: Springer; 2002. 4

5

28. Cools M. Inter- and Intraday Variability of Flemish Travel Behavior. 2009; 6

7

29. Lynch K. The Image of the City. The MIT Press; 1960. 8

9

30. Schuitema G, Steg L, Vlek C. Prijsbeleid: effectief en aanvaardbaar voor veranderen van 10

autogebruik? Colloquium vervoersplanologisch speurwerk: no pay, no queue? 11

Oplossingen voor bereikbaarheidsproblemen in steden 2003;:385-400. 12

13

31. Bargh JA, Chartrand TL. The Unbearable Automaticity of Being. American Psychologist 14

1999 Jul;54(7):462-479. 15

16

32. Miermans W, Van Moerkerke B. Mobiel Vlaanderen - Mobiliteitsconvenant [Internet]. 17

Mobiel Vlaanderen 2005;[cited 2010 Apr 26] Available from: 18

http://www.mobielvlaanderen.be/convenants/artikel.php?nav=0&id=200 19

20

33. Shiftan Y, Golani A. Effect of Auto Restraint on Travel Behavior. Transportation 21

Research Record: Journal of the Transportation Research Board 2005 Jan;1932(-1):156-22

163. 23